← Back to Skills Marketplace
146
Downloads
1
Stars
0
Active Installs
4
Versions
Install in OpenClaw
/install project-trident
Description
Four-tier persistent memory architecture for OpenClaw agents. Implements LCM-backed durability, hierarchical .md file organization, agentic signal routing, a...
Usage Guidance
What to check before installing:
- Expect network/cloud usage: SKILL.md and reference scripts assume use of Anthropic/OpenAI/Gemini, Qdrant/FalkorDB, Ollama, and Git remotes. The manifest does not list required API keys or credentials — ask the author to declare them or refuse to provide secrets until clarified.
- Inspect scripts before running: open scripts/migrate-existing-memory.sh and scripts/template-integrity-check.sh and any Python indexing code. Look for unexpected network calls, exec/curl|sh pipelines, or commands that delete or push files.
- Dry-run migration: always run ./migrate-existing-memory.sh --dry-run first and verify backups in memory/migration-backup/ before allowing changes.
- Avoid pushing private memory remotely by default: do not add a git remote or run git push until you are sure the backup cron only commits safe files and .gitignore protects secrets (SOUL.md, USER.md, .env, etc.). Consider keeping backups local or in a private, offline store.
- Limit credentials and scope: If you enable semantic recall or Qdrant cloud, create dedicated, minimally scoped API keys and restrict network egress; prefer local Ollama or local Qdrant binaries for sensitive deployments.
- Prefer isolation: consider running Trident in a dedicated user account or container/VM so cron jobs and file reads are limited to the Trident workspace path.
If the author cannot explain and update the manifest to declare required env vars/credentials and clarify what is sent to external services, treat this skill cautiously (suspicious).
Capability Analysis
Type: OpenClaw Skill
Name: project-trident
Version: 2.0.1
Project Trident is a comprehensive memory management framework for OpenClaw agents, utilizing a tiered architecture to ensure data persistence through SQLite, hierarchical Markdown files, and automated backups. The bundle includes administrative shell scripts for workspace migration (migrate-existing-memory.sh) and SHA256-based integrity verification (template-integrity-check.sh) to protect the AI agent's routing instructions from tampering. While the system requires significant permissions for file manipulation and cron job scheduling, all behaviors are transparently documented and strictly aligned with the stated goals of providing agent continuity and disaster recovery without evidence of malicious intent or data exfiltration.
Capability Tags
Capability Assessment
Purpose & Capability
Name/description describe a local four-tier memory system. That's coherent with included scripts and .md file organization. However SKILL.md and reference files repeatedly instruct use of external LLM APIs (Anthropic, OpenAI, Gemini), Qdrant/FalkorDB cloud services, and Git push workflows — yet the skill's metadata declares no required environment variables or primary credential. Requiring cloud API keys, SSH keys, or external accounts is plausible for the described optional features, but the manifest should declare them. The omission is an incoherence: the capability expects secrets and network access that the registry metadata doesn't advertise.
Instruction Scope
The instructions tell the agent (and the user) to enable plugins in openclaw.json, create cron jobs that run every 5–30 minutes, read and write many workspace .md files, index and embed those files, and optionally push memory to remote Git repositories or cloud vector stores. Those actions are within the skill's purpose (managing agent memory) but they also grant broad access to the user's workspace and could exfiltrate sensitive text if a cloud backend or git remote is configured. The instructions also call out a Python indexing script that uses the OpenAI client and a Qdrant client without declaring the required keys.
Install Mechanism
No install spec in the manifest (instruction-only), which reduces direct installer risk. The documentation does recommend downloading third-party binaries (Qdrant/FalkorDB releases), using curl|sh for Docker/Ollama installs, and running system package commands. Those are common but carry the usual risks of executing remote install scripts; the manifest does not centralize or vouch for these downloads.
Credentials
The skill's runtime content clearly expects credentials and network endpoints: OpenAI/embeddings, Anthropic model access, Gemini API keys, Qdrant/FalkorDB cloud credentials, Git SSH keys (for pushing backups), and possibly Ollama local installs. None of these are declared in requires.env or primaryEnv. That mismatch means users may unknowingly wire secrets into system cron jobs or agent prompts; the skill also suggests pushing memory (which can include private data) to remote Git hosts.
Persistence & Privilege
always:false and model-invocation allowed (defaults) — normal for an agent skill. The skill instructs creating persistent cron jobs and modifying OpenClaw config (enabling the 'lossless-claw' plugin), which grants it lasting operational presence and scheduled autonomous activity. This is reasonable for a memory system but increases blast radius if external connectors (cloud APIs, git remote) are added — the manifest should be explicit about that.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install project-trident - After installation, invoke the skill by name or use
/project-trident - Provide required inputs per the skill's parameter spec and get structured output
Version History
v2.0.1
v2.0.1: Four-tier architecture with Layer 2 backup subsystem. Integrated Heartbeat signal router. No Docker required (Trident Lite). Full Windows/Mac/Linux support. Interactive migration script. SHA256 template integrity. Cost calculator with 5 profiles. Complete disaster recovery procedures. Ready for mass adoption.
v2.0.0
v2.0: Mass-adoption release. No Docker required (Trident Lite is default). Full Windows/Mac/Linux support with platform-specific guides. Interactive migration script for existing agents (dry-run + auto-backup). SHA256 template integrity check with audit logging. Cost calculator with 5 profiles including Gemini Flash and Ollama. Deployment guide restructured into three tracks. Security-hardened from v1.
v1.2.0
v1.2.0: Generalized deployment — removed infrastructure-specific refs (Hostinger/VPS/Docker-only). Now deployment-agnostic: local-first, cloud optional. Core: Layer 0 (LCM), Layer 0.5 (signal router), Layer 1 (hierarchical .md buckets). Optional: Qdrant (Docker/cloud/binary), FalkorDB (Docker/cloud/Redis). Passes ClawHub security review.
v1.1.0
- Improved documentation: SKILL.md now details Trident's multi-phase architecture and use cases.
- Clarifies phase progression (1–8) and bridges connection between SQLite durability, .md buckets, agentic routing, and cloud backup.
- New sections provide clear rationale, cost estimates, daily workflows, and detailed implementation checklists for each phase.
- Update highlights how the system solves context continuity and "blank spots" for autonomous agents.
Metadata
Frequently Asked Questions
What is Project Trident?
Four-tier persistent memory architecture for OpenClaw agents. Implements LCM-backed durability, hierarchical .md file organization, agentic signal routing, a... It is an AI Agent Skill for Claude Code / OpenClaw, with 146 downloads so far.
How do I install Project Trident?
Run "/install project-trident" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Project Trident free?
Yes, Project Trident is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does Project Trident support?
Project Trident is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Project Trident?
It is built and maintained by Shiva&G (@shivaclaw); the current version is v2.0.1.
More Skills